Blending Optimal Control and Biologically Plausible Learning for Noise-Robust Physical Neural Networks

TL;DR

This paper introduces a novel training method for physical neural networks that combines optimal control and biologically plausible learning, significantly reducing training costs and enhancing noise robustness, verified in optoelectronic systems.

Contribution

It presents a new training approach merging optimal control with feedback alignment, enabling efficient, noise-robust training of physical neural networks.

Findings

Reduces training time for physical neural networks.

Achieves robustness against measurement errors and noise.

Validated in optoelectronic delay systems.

Abstract

The rapidly increasing computational demands for artificial intelligence (AI) have spurred the exploration of computing principles beyond conventional digital computers. Physical neural networks (PNNs) offer efficient neuromorphic information processing by harnessing the innate computational power of physical processes; however, training their weight parameters is computationally expensive. We propose a training approach for substantially reducing this training cost. Our training approach merges an optimal control method for continuous-time dynamical systems with a biologically plausible training method--direct feedback alignment. In addition to the reduction of training time, this approach achieves robust processing even under measurement errors and noise without requiring detailed system information. The effectiveness was numerically and experimentally verified in an optoelectronic delay system. Our approach significantly extends the range of physical systems practically usable as PNNs.